PayMCP

Converts any MCP tool into a paid endpoint using a lightweight Python or TypeScript decorator that intercepts tool invocations, validates payment credentials, and gates execution. The decorator pattern wraps the original tool function without modifying its signature, injecting payment validation logic at runtime before the tool executes. Supports multiple payment providers through a pluggable backend architecture.

Provides a unified interface for integrating multiple payment backends (Stripe, custom HTTP endpoints, etc.) through a pluggable provider pattern. The abstraction decouples tool payment logic from specific payment provider implementations, allowing developers to swap providers or support multiple providers simultaneously without changing tool code. Implements provider-agnostic validation and error handling.

Manages authentication credentials and payment tokens for tool invocations, validating that incoming requests include valid payment authorization before tool execution. Implements credential extraction from MCP request context, token validation against payment provider, and credential caching to reduce provider API calls. Supports both API key and OAuth token patterns.

Automatically captures payment-related events (authorization attempts, successes, failures, balance changes) and generates structured audit logs for compliance and debugging. Logs include timestamp, user ID, tool ID, payment status, provider response, and error details. Supports custom log handlers for integration with external logging systems (CloudWatch, Datadog, etc.).

Enforces usage quotas and rate limits on paid tools based on user subscription tier or payment status, preventing abuse and ensuring fair resource allocation. Implements quota tracking (calls per minute/hour/day), tier-based limits (free tier: 10 calls/day, pro tier: 1000 calls/day), and quota reset scheduling. Integrates with payment provider to determine user tier and remaining quota.

Implements configurable error handling for payment provider failures, including retry strategies (exponential backoff, jitter), fallback behaviors (deny access, allow with deferred payment, etc.), and detailed error reporting. Distinguishes between transient failures (network timeout, provider temporarily unavailable) and permanent failures (invalid credentials, insufficient balance) to apply appropriate retry logic.

Provides identical decorator-based payment gating API in both Python and TypeScript, allowing developers to use the same patterns regardless of implementation language. Maintains feature parity between implementations (same decorator syntax, same provider abstraction, same configuration format) while using language-native patterns (Python decorators, TypeScript decorators). Shared documentation and examples work across both languages.

Integrates directly with the MCP protocol layer to extract payment credentials and user context from MCP request metadata, without requiring explicit parameter passing in tool signatures. Implements MCP context parsing to retrieve user ID, API key, subscription tier, and other payment-relevant metadata from MCP request headers or custom context fields. Operates transparently to tool implementations.